Docker compose 원하는 log만 출력하기

기본적으로 

docker-compose up

으로 container들을 실행하면 모든 서비스에 대해 log가 다 출력된다.

 

1. docker compose detached 모드로 실행

docker-compose up -d

output:

Creating network "listen-attend-and-spell-pytorch_default" with the default driver
Creating listen-attend-and-spell-pytorch_train_1       ... done
Creating listen-attend-and-spell-pytorch_tensorboard_1 ... done

생성되며 백그라운드에서 실행된다.

 

2. 특정 서비스에 대해서만 log 출력

docker-compose logs -t -f train

output:

Attaching to listen-attend-and-spell-pytorch_train_1

Attach되며 log가 출력되기 시작한다.

-f means you follow the log output and the -t option gives you timestamps

-f는 service의 log output을 따른다는 의미고, -t는 log에다가 timestamps를 달아준다.

 

output:

train_1        | 2020-04-16T07:57:17.230959517Z [2020-04-16 07:57:17,230 train_timit.py:96 - <module>()] epoch: 0, global step:    924, loss: 3.6739, cer: 0.8311, elapsed: 6.14m 0.10h
train_1        | 2020-04-16T07:57:20.965151763Z [2020-04-16 07:57:20,965 train_timit.py:116 - <module>()] epoch: 0, global step:    924, loss: 3.6617, cer: 0.8159, elapsed: 6.20m 0.10h
train_1        | 2020-04-16T08:03:32.005941448Z [2020-04-16 08:03:32,005 train_timit.py:96 - <module>()] epoch: 1, global step:   1848, loss: 3.7034, cer: 0.8128, elapsed: 6.18m 0.21h
train_1        | 2020-04-16T08:03:35.761454523Z [2020-04-16 08:03:35,761 train_timit.py:116 - <module>()] epoch: 1, global step:   1848, loss: 3.6873, cer: 0.8101, elapsed: 6.25m 0.21h

이미 내 python 코드 log에 timestamp가 존재하는데 중복이 되었다....

 

3. 결론

Docker container를 2개 이상 사용한다면 docker compose는 필수라고 생각한다.

 

출처:

https://stackoverflow.com/questions/37195222/how-to-view-log-output-using-docker-compose-run